Twitter’s Negligence in Removing IRA-Linked Accounts

Unhackthevote November 22, 2019

Share:

In August 2018, Twitter announced that it had identified and suspended approximately 3,000 IRA-linked Twitter accounts. (Subsequently, they pulled 228 names from that list as Venezuelan-linked rather than Russian-linked.)

Information from the IRA-linked Twitter accounts was given to the FBI for review and some of the data published publicly. (Note: Twitter previously did not require an email address to access the information—but it does now.)

These are the accounts referenced in Mueller’s “Report on the Investigation into Russian Interference in the 2016 Presidential Election.” (Volume 1, PDF).

With this purge of IRA-linked accounts, one might get the impression that Twitter was diligently tracking IRA-linked accounts and trying to be helpful with the ongoing investigation into Russian election interference. But that impression would be faulty. On the contrary, Twitter appears to be doing as little as possible to identify and suspend IRA-linked accounts.

The same day Twitter released the data for the IRA-linked Twitter accounts, a group of researchers spotted a part of a bot network in Twitter’s data dump. That network led them to hundreds more IRA-linked accounts left unidentified by Twitter. At the time of the discover, the accounts were still active on the platform.

But they are not the only ones who spotted more bots connected to the IRA-linked accounts Twitter published. After scanning the IRA-linked Twitter data for less than 30 minutes, our lead bot researcher spotted another bot-group pattern.

Following the pattern, we found over 1,000 active, IRA-linked Twitter bots before we stopped counting.

A Basic Bot Fact

One of the most basic facts about fake Twitter accounts is that they are often created in groups of similarly structured accounts. For example, here are a few accounts from one bot group. Note the similarities: no background pics, short phrase for a profile description, similar following and follower counts.

Did Twitter Ignore an IRA-Linked Account Pattern?

What happens when we apply what we know about bot group patterns to the IRA-linked accounts Twitter identified and suspended?

Twitter’s data dump (csv) appears to be little help because it’s mostly consists of cells filled with indecipherable alpha-numeric strings.

Fake Account Mad Libs

Upon closer inspection, however, there is a discernible pattern in some of the profile descriptions. This is a pattern we’ve seen before—in a bot group we had spotted in our ongoing study of fake Twitter accounts.

We refer to this pattern as “Mad Libs.” Like the childhood game, this fake account group mixes distinctive, descriptive words with specific nouns, resulting in peculiar phraseology that they use in their profile descriptions. This phraseology is repeated across the accounts within the group.

Here are some examples from the IRA-linked accounts that Twitter identified and suspended. Notice how certain nouns and descriptive words are used repeatedly in a mix-and-match fashion across the accounts.

And here is a sample from a still-active group of fake Twitter accounts with identical phraseology. This strongly suggests that these accounts are also IRA-linked.

Where are Twitter’s fancy algorithms?

We have documented over a thousand of these still-active, fake accounts and similar, interconnected accounts. But this makes us wonder: if we can find these accounts with nothing more than our eyes (and for free), how did Twitter miss them?

Granted, like all bot-group patterns on Twitter, the “Mad Libs” pattern stands out when looking at the accounts together. But, it becomes difficult to see as the bot group is spread out among hundreds of millions of other Twitter accounts—at least for regular users. But what about Twitter?

Was Twitter—with its fancy algorithms and billions of dollars—truly incapable of seeing such a simple pattern and tracking down fake accounts that mimic it? Or is something else going on? Whatever the answer is, it seems fair to say that, at best, Twitter has been negligent in rooting out Russia’s social media army—an army that influenced the 2016 U.S. Presidential election; an army that continues its assault on America.

Written by Unhackthevote